Our relationship with other intelligences, such as animals, is complex. AI and AI’s physical manifestation in robots is complex too.
Kate Darling, a researcher at MIT Media Lab and an expert in robot ethics and policy, writes that we need to rethink human-robot metaphors to make sure we can capture their full potential. Our constant subconscious comparison of robots to humans can lead to inappropriate and suboptimal outcomes in design, use, ethics, and law. If we insist on thinking of robots as humanoids we are blind to how humans actually interact with them. Instead we should think of relationships with robots as more like our relationships with animals.
Animals have been our tools and companions for centuries. We don’t expect animals to have our intelligence—we recognize their role in supplementing our capabilities, whether it be because they bear large loads, detect scent, have incredible sight, fly or find. We also find comfort, companionship, calmness, and healing in the presence of animals. We project ourselves onto them, anthropomorphize their behavior, grieve for them when they die, and imagine ourselves better because of them.
It is a function of history, culture, and the usefulness of the animal. Pigs use tools, understand symbolic language and mirrors, can be house trained, wag their tails when happy, control their environment, and are empathetic. On a factory farm, pigs will spend most of their adult life in a cage measuring two feet by six and half feet. Dogs and pigs are not that different, yet depending on where you are, they both taste good. In the USA, eating dogs and cats was only made illegal in all states in 2018 through the Dog and Cat Meat Trade Prohibition Act.
Compare this to how, on any given day, there are 75 million pigs in factory farms in the US and more than 120 million pigs are killed for food every year. Human societies draw borders around our relationships with animals in many ways, as we will with robots.
We need to build new mental models so that we can realize the partnership opportunities that robots—and by extension, all AI—will provide.
Robots are a very special kind of AI because their ability to move triggers our psychology in different ways than the AI we meet through screens. We are acutely sensitive to movement because we perceive aliveness which alerts us to danger. Robots amplify everything. If we can imagine new minds by imagining new robots, we can see how potent machine intelligence can be.
Humans project their own minds onto robots, so robots also offer us a reflection of our humanity. When someone kicks a robot, other people feel uncomfortable with mistreatment. There is some evidence that abusing a robot is psychologically like abusing an animal—it trains our cruelty muscles [Darling].
People fall in love with their robots. Soldiers have been known to enter a battlefield to retrieve a fallen robot comrade. Empathy makes us feel sorry for a Roomba that gets stuck in a corner. Some believe that robots will take care of us when we are old, lonely, and infirm. The tangibility of a robot-human relationship and how strongly it can evoke our emotions can lead us to forget why the robot is there in the first place and what it is there to do.
Emotional ties to robots are a double-edged sword. Anthropomorphism could stop us seeing other ways to use machines but also feel good about having robots in our lives, performing tasks that humans used to perform.
Perhaps there are advantages because machines do not care. For now, machines feel no suffering. Machines can be good partners in holding you accountable to difficult choices. They feel no sense of dread and won’t be anxious about what people think of them or whether they are about to be fired.
Machines can make us feel free. We don’t need to care if we’ve hurt a machine’s feelings. Many people prefer talking to a machine because they feel less self-conscious. They can feel a conversation with a machine doctor, for example, is more private.
We can run experiments on a machine mind without being worried about whether we should or not. We can test what machines think as much as we want. It doesn’t get tired of being asked nor does it change its answer in a fickle, mischievous way just to get a reaction. A machine doesn’t need to poke, tease, or use sarcasm.